It is the year 2525. Earth and Mars are in a state of cold war against each other. Both planets import uranium from Ceres. Ceres' government is pro-Earth…for now. Earth needs to know how likely it is the Ceres government will be overthrown and replaced by a pro-Mars regime.

Earth creates a prediction market on the blockchain. Earth seeds the prediction market by buying shares that pay out if there is a coup. This lopsided approach prevents Earth from directly subsidizing a coup it doesn't want.

Earth must provide lots of liquidity because if Earth provides too little liquidity then Mars can cheaply influence Earth's decisions by manipulating the prediction market's prices.

Ceres' allegiance is zero-sum. Mars wants there to be a coup because an anti-Earth regime is a pro-Mars regime and vice versa. Mars can subsidize a coup by buying shares that pay out if there is not a coup. Mars' investment simultaneously acts as an investment in a coup and insurance against the coup failing to materialize.

Suppose Mars values a coup at 5 million Dogecoin. Mars buys shares that pay out 5 million Dogecoin if there is not a coup. Mars' investment subsidizes a coup because entrepreneurial space pirates can become Mars' counterparty and use their expected profits to fund the conquest of Ceres. Even more efficiently, the Ceres government could buy the shares itself and then change its foreign policy to be pro-Mars, thereby realizing a risk-free profit without any violence at all (while also discouraging space pirates).

But from Mars' perspective it no longer matters if there is a coup because if there is a coup then Mars' windfall of 5 million Dogecoin exactly neutralizes the value it gains from a coup.

The less likely a coup is, the more Mars stands to gain by buying insurance. Suppose the prior implied probability of a regime change is 0.20. Mars can buy its shares for 1 million Dogecoin, pocketing a risk-free net utility equivalent to 4 million Dogecoin. If, alternatively, the prior implied probability of a regime change is 0.80 then Mars pockets a risk-free profit of only 1 million Dogecoin.

Where does this free money come from? Opportunity cost. The alternative to buying blockchain prediction market insurance is for Mars to send its own marines to overthrow the Ceres government. Mars landing troops on Ceres is only profitable if Mars does not buy insurance because insurance directly neutralizes the profits of a coup on Ceres. It might be that deploying marines is cheaper than hiring spaces pirates. But even if hiring marines is technically cheaper, operating through the prediction market provides a level of deniability overt operations lack. After all, Mars' enthusiastic participation in a blockchain prediction market created by Earth can hardly be considered a breach of the peace.

Mars' strategy requires the prediction market contains sufficient liquidity—which Earth conveniently provided. Thus, the strategy of buying shares that pay out in outcomes you dislike I outlined here fails when you have enemies because your enemies can leverage the prediction market's liquidity to their advantage.

New to LessWrong?

New Comment
21 comments, sorted by Click to highlight new comments since: Today at 2:17 PM

Mars' investment subsidizes a coup because entrepreneurial space pirates can become Mars' counterparty and use their expected profits to fund the conquest of Ceres.

 

Ok, so what happened is Mars bet 5 million dodgecoin there would not be a coup.  So if there is a coup, those who bet for a coup gain 5 million dodgecoin?  Just trying to understand this nested statement.  So now the space pirates have a speculative venture : conquer Ceres.  How do they get investors to fund their expedition and pay for all the robotic marines and munitions and ships they need?  Is there a blockchain solution to lending for this?

For that matter, don't the space pirates have to have 5 million dodgecoin in liquidity to make the bet for the expedition and also another 5 million to pay for it?  It seems like this situation just offers a chance for them to get paid but not actually pay the 'cash' they need to buy weapons off the black market.

Ok, so what happened is Mars bet 5 million dodgecoin there would not be a coup. So if there is a coup, those who bet for a coup gain 5 million dodgecoin? Just trying to understand this nested statement…It seems like this situation just offers a chance for them to get paid but not actually pay the 'cash' they need to buy weapons off the black market.

You are correct. The system guarantees space pirates will get paid if they topple the Ceres government. It does not provide them with upfront liquidity. After all, there is no way for Mars to know that this batch of alien ruffians can be trusted. (Especially because if the space pirates were paid upfront then they would no longer be incentivized to conquer Ceres.)

How the space pirates turn an expected windfall of 5 million Dogecoin into immediate liquidity is left unspecified. It depends on local customs and conditions. But a probable future windfall can usually be turned into upfront cash (for a fee). There may be a blockchain solution or they might just to go a local investor with whom they have established trust (but who does not need connections to the Martian government).

Mars bet 5 million dodgecoin

'dogecoin', not 'dodgecoin'

I wonder if this is a direct example of Goodheart's Law.  When a measure (prediction) is used as a control (to make decisions on), it ceases to be a good measure.

Are there formal models of the behavior of prediction markets like this? Some questions that such a theory might answer:

  • Is there an equivalence between, say, "I am a bettor with no stakes in the matter, and believe there is a 10% chance of a coup", and "I am the Mars government and my utility function prefers 'coup' to 'not-coup' at 10-to-1"? In both cases, it seems relevant that the agent only has a finite money supply: if the bettor only has $1, the profit they can make and the amount they can move the market is limited, and if Mars "only" stands to gain $5 million from the coup then they're not willing to lose more than $5 million in the market to make it happen.
  • In a group of pure bettors, what's the relationship between their beliefs, their money supply, and at what price the market will stabilize? I'm assuming you'd model the bettors as obeying the Kelly criterion here. If bettors can learn from how other bettors bet, what are the incentives for betting early vs. late? I imagine this has been extensively studied in economics?
  • If you want to subsidize a market, are there results relating how much you need to subsidize to elicit a certain amount of betting, given other assumptions?

If Mars values a coup at 5M, and Earth values a coup at -5M, Earth can buy contracts to win 5M if there is a coup, Mars can buy contracts to win 5M if there isn't a coup, and they can both cancel their clandestine programs on Ceres, making the interaction positive-sum.

...Not sure that actually works, but it's an interesting thought.

It works (in theory (assuming rational, liquid markets)) if you assume that they both value risk-reduction and that they don't care about each others' finances. There is no way to make the interaction positive sum if we assume the Earth-Mars relationship is a priori zero-sum, since that would be a contradiction.

Note that the cancellation of Earth's and Mars' clandestine programs on Ceres does not actually mean Ceres is free of foreign influence. It just means that Ceres is free of direct foreign influence. Earth and Mars have offloaded their operations to privateers and/or indirect bribes to the government of Ceres.

What if Earth's prediction market only allows Earth's government to buy the "pays out if coups happen" shares, and only allows the counterparties to buy the "pays out if coups don't happen" shares? Couldn't this eliminate the space pirate problem?

If Earth builds its prediction market on the blockchain then other parties could create derivative smart contracts to evade Earth's restrictions. Earth would have to limit prediction market trading to its domain of territorial influence and carefully monitor all of the traders' financial dealings, which would weaken the power of the prediction market.

I thought the whole blockchain point was just a funny story, not some serious proposal? Outside of dodging regulations, I don't really see the reason why you'd have a blockchain-based prediction market. It seems like it would be much easier to just store it in a big database.

As for the issue of limiting trading, if you are running the prediction market on your own servers with your own code, it would be trivial to prevent participants to trade with each other, simply by not implementing it or anything that can be used for it (like exchanges of contracts). This would also seem to prevent setting up derivative contracts because people have no way of enforcing them.

The Earth/Mars/Ceres stuff was for a fun story (and to keep the example grounded while avoiding unnecessary real-world politics) but the blockchain is an important ingredient. The blockchain is not there to dodge regulation. Earth doesn't need to dodge regulations. It is a sovereign state. Earth writes the regulations. It can do whatever it wants, legally-speaking.

The purpose of using a blockchain is precisely to commit Earth into not doing the shenanigans you propose. If Earth plans to interfere with the prediction market then you no longer have a free (as in markets) prediction market.

There are ways of implementing contracts based off of Earth's market without a blockchain (especially if Earth's prediction market prices are public knowledge) but they are messy. Using a blockchain keeps the thought experiment simpler by allowing arbitragers to just program derivative smart contracts.

The purpose of using a blockchain is precisely to commit Earth into not doing the shenanigans you propose. If Earth plans to interfere with the prediction market then you no longer have a free (as in markets) prediction market.

But why not just do the shenanigans I propose?

Also it doesn't seem like it's reasonable to call them interference. It's not that Earth changes the payouts or anything. It just limits the trading.

There are ways of implementing contracts based off of Earth's market without a blockchain (especially if Earth's prediction market prices are public knowledge) but they are messy. Using a blockchain keeps the thought experiment simpler by allowing arbitragers to just program derivative smart contracts.

Wouldn't these require basically setting up a separate prediction market? Which would mean that your enemies are not using your market, but are instead using this separate market.

But why not just do the shenanigans I propose?

If Earth had a totally insulated economy (like North Korea) then you could do it. Otherwise, your shenanigans amount to banning your traders from hedging their risks. Prediction markets work because people arbitrage away risk. Arbitrage and hedging are two sides of the same coin. (See the Chapter 5 of this book for a precise mathematical formulation.)

To put the situation another way, it would do no good for Earth to just ban traders from hedging away the risk on the Earth servers. Earth would have to ban traders from hedging away their risk on all markets, period. That implies you're insulating Earth's prediction market from the rest of the Solar System's economy—and you can't do that while incentivizing experts to use the prediction market via profits exchangeable for interplanetary Dogecoin.

Wouldn't these require basically setting up a separate prediction market? Which would mean that your enemies are not using your market, but are instead using this separate market.

Doesn't matter. The separate prediction market is funded indirectly by the liquidity provided by the first prediction market. The separation of markets is just a legal cover.

Prediction markets work because people arbitrage away risk.

This sounds wrong. Prediction markets work because you run them with generous subsidies to incentivize participation; risk-based arbitrage is not a recommended way to fund the markets.

Doesn't matter. The separate prediction market is funded indirectly by the liquidity provided by the first prediction market. The separation of markets is just a legal cover.

Just so I understand: What you have in mind is something like the following?

  • Earth creates a prediction market, betting hard that Ceres is going to have a coup
  • Mars creates a corresponding prediction market, without betting anything
  • Some trader, who I'll call Ivan, offers a bet that Ceres is going to have a coup on Mars's market
  • If anyone takes him up on it, Ivan will buy a corresponding contract from Earth's market, pocketing a small profit and eliminating all risk from his perspective
  • This generates liquidity in Mars's market, about as much liquidity as there is on Earth's market

And I guess I understand you so far. What happens then? Is it something like:

  • Initially, maybe the Earth/Mars relations are rather cold, so nobody expects a coup; therefore they bet it all the way up to 95% probability that there will be no coup.
  • But this now creates a strong incentive to make new bets that Ceres is going to have a coup and then make Ceres have that coup, e.g. Hook the Space Pirate sell contracts in the Martian market and then start a coup on Ceres.

I feel like at least this point of your original post fails though:

But even if hiring marines is technically cheaper, operating through the prediction market provides a level of deniability overt operations lack. After all, Mars' enthusiastic participation in a blockchain prediction market created by Earth can hardly be considered a breach of the peace.

Mars buys shares that pay out 5 million Dogecoin if there is not a coup

Suppose the prior implied probability of a regime change is 0.20. Mars can buy its shares for 1 million Dogecoin, pocketing a risk-free net utility equivalent to 4 million Dogecoin.

I'm confused - if the prior probability of a coup is 20% and Mars is buying shares that pay out if no coup, Mars would pay 4M?

A prior implied probability of 20% that there will be a coup means Mars would pay 0.2 for shares that pay out 1.0 if there is a coup. Multiply both sides of the equation by M (1 million). It costs Mars 0.2M to buy shares that pay out 1M if there is a coup. Multiply both sides of the equation by another 5. It costs Mars 1M to buy shares that pay out 5M if there is a coup.

If Mars pays 1M to buy shares that pay out 5M if there is a coup then Mars pockets 5M − 1M = 4M.

Yes, but you said they're buying the no-coup shares, which subsidizes a coup. Article contradicts itself. 

Thanks for the interesting question.  So now the other question: it seems that these payments make these prediction markets causal, they themselves affect the outcome of their own predictions.  

This seems like a bug.  What about a prediction market based on 'reputation', where you start with little reputation and each accurate prediction, made in advance of the event happening and signed to the blockchain, increases reputation.  You could also in theory use a similar kind of market for 'observation', as in, an observer who is a fair observer of events would record what they observed and gain reputation while 'alternative facts' biased observers lose reputation.  Presumably the reputation updates would require some kind of inference to be performed automatically. 

It has occurred to me that to make this work, both "observers" and "predictors" must be AI systems with fixed source code with their hashes committed to the blockchain.  This makes them deterministic and this determinism can be checked by all chain members.  

Anyways regardless of how we do it, this seems like a bug.  The reason the "market price" is a fair price is because if it were consistently unfair, someone could make money off trades with their knowledge of the consistent unfairness, and after they gain sufficiently money, their 'votes' as a market participant grow in power until well informed participants effectively control the market price.  (IRL example : hedge funds).  

Presumably a market with 'reputation' or some other meta value that can be gained or lost would have this same convergence.  

So then how do you reward market participants to give them an incentive to gain reputation.  What if...you could put predictions/observations onto the blockchain in encrypted form and later put on the blockchain the key.  Then you could get paid as a high reputation observer/predictor to share your observations/predictions in advance with paying customers. 

The mechanics here are pretty similar to assassination markets.

Built in to the premise of the question is that Earth values Ceres' loyalty more than Mars (otherwise Mars presumably wouldn't be able to use the prediction market as a hedge). I agree (I think?) that the prediction market lets you outsource the work to space pirates, but it seems also necessary to point out that at Earth is sending at least as much money as a counter force (encouraging 3rd parties to keep Ceres loyal).

[+][comment deleted]2y10